JP2834130B2 - How to check recognition data - Google Patents

Description

The present invention relates to a method and an apparatus for efficiently confirming the result of automatic recognition from drawings and maps, and particularly to a method for recognizing a large-size drawing as a CRT. To a method that enables confirmation and editing in a limited area.

[Conventional technology]

Conventionally, the document “Work-handling report of handwritten design drawing automatic input technology” (“Equipment drawing management system”,
(Kanachika, Watanabe, Tanaka) (hereinafter referred to as Reference 1). A method for inputting drawings and documents and confirming the result of automatic recognition is described in the document “Handwritten design drawing automatic input technology work shop. Report, 1980 "(" Automatic input of handwritten device system diagram "Ishii, Ito) (hereinafter referred to as Reference 2)
In, it is discussed. In Document 2, (1) the original image and the recognition result are superimposed and displayed on the display, (2)
A method has been proposed in which a confirmation signal is displayed on another screen of the multi-window to facilitate visual confirmation, and (3) a symbol is designated, and the symbol is searched for and displayed on the drawing. I have. In the method (1), when the display plane is limited to the CRT, as in the case of confirming and correcting on a workstation or personal computer, if the original drawing is to be displayed, the original drawing is reduced or displayed. Must be limited to a part of the original drawing.

In (2) and (3), a method of enlarging a display plane using a multi-window function or limiting recognition results to be displayed at a time in order to display a check result in a limited space to a certain extent. Has been proposed. On the other hand, the document “PRL 82-6
6,1982 ”(“ Interactive inspection system of image processing algorithm and its application to meteorological map processing algorithm ”by Nimura et al.), When a plurality of types of symbols are recognized in the method (1) of Reference 2, There has been proposed a method of displaying each symbol in a different color.

[Problems to be solved by the invention]

The method of displaying the recognition result on the image data of the original drawing in Literature 2 does not consider the space to be displayed, and furthermore, the recognition is an operation on the drawing. This is a search for the recognized symbol, which causes omission of confirmation. In the methods (2) and (3), the recognition result to be displayed is limited to the symbol to be processed. By displaying a plurality of symbols at the same time, it is possible to perform recognition by comparison with other recognition results. Can not. Furthermore, (2), (3)
The method requires a display device having a multi-window function, and the use environment is limited.

The object of the present invention is to solve the above-mentioned problems, and when the target drawing is large or complicated, in a limited space,
It is to be able to confirm and correct work efficiency and quality.

In the past, the confirmation operation was performed on the read drawing, whereas in the present invention, the operation target was an image that extracted only the information necessary for the recognition operation from the drawing, so the confirmation operation was performed in a limited space. Enable operations.

Conventionally, recognition results were confirmed by checking the recognition symbols individually to determine whether the confirmation results were correct. Checking the result increases the work efficiency instead of the simpler operation of searching for different symbols from the same kind of symbols.

Furthermore, in the conventional method, a search for a symbol that was incorrectly recognized was performed in a two-dimensional area of a drawing. Therefore, it is necessary to search for symbols in one-dimensionally arranged ones, which has the effect of preventing omission of confirmation and enhances the quality of confirmation work.

[Means for solving the problem]

The above object is to cut out an image from the drawing to be recognized around the center coordinates of the recognized symbol and create a table arranged for each symbol (hereinafter, this table is referred to as a check sheet).
The recognition result is confirmed by using a check sheet, and as a result of the confirmation, a symbol determined to be erroneously recognized is deleted and corrected from a list obtained as a recognition result.

[Action]

In the recognition result confirmation method proposed in the present invention, a check sheet in which peripheral images of recognition symbols are cut out and arranged from a list of recognition results is created and used for confirmation, so that the confirmation result is confirmed. Sometimes, it is not necessary to look at the drawing, and even when the display space is limited, it is possible to display a plurality of symbols in an easily viewable size, thereby facilitating the confirmation operation. Since the results of recognizing the same type of symbol are displayed at the same time, the operation of finding a misrecognized symbol is an operation of searching for a different symbol from the same type of symbol, and the operation is simple.

In addition, the symbol displayed on the check sheet is associated with the list of recognition results in the temporary memory of the computer stored when the automatic recognition was performed. It is also easy to correct what is determined to be.

Further, for a drawing or the like containing confidential information, a recognition method using a check sheet is adopted, so that the checking operator does not need to directly contact the drawing, which is useful for maintaining confidentiality.

〔Example〕

Hereinafter, embodiments of the present invention will be described with reference to the drawings. First
The diagram is proposed in the present invention. It is an example which shows the specification of a check sheet. FIG. 2 shows a basic configuration necessary for realizing the present invention. FIG. 2 (a) is a configuration diagram until a check sheet is created, and FIG. 2 (b) is a configuration diagram until a recognition result is confirmed and corrected using the check sheet. Reference numeral 11 denotes image data of a drawing to be automatically recognized, and reference numeral 12 denotes a list obtained as a result of the recognition. Reference numeral 13 denotes a processing device having an image processing function, and creates 14 check sheets from the data of 11 and 12. Reference numeral 15 denotes a person who performs a checking operation, and inputs a correction instruction to the processing unit 16 while viewing the created check sheet. Reference numeral 16 denotes a processing device having correction processing means for internally storing a list of twelve recognition results for an input. FIG. 3 is a configuration diagram of a system for realizing the present invention. Third
In the drawing, reference numeral 21 denotes a drawing to be automatically recognized, such as the drawing shown in FIG. 6A, and reference numeral 22 denotes an image reader which reads the drawing as image data. 23 is a memory for temporarily storing a result of automatic recognition of a drawing to be processed and data in the middle of processing, 24 is a file device for storing a list of read image data and recognition results, In the figure, reference numeral 25 denotes a processing device such as a CPU of a microprocessor or a minicomputer, which has 13 image processing functions and 16 correction processing means. 28 is a CRT for displaying image data and vector data, and 29 is a coordinate input device for pointing a position on the CRT.

In such a configuration, the entire flow of the present invention is changed to the fourth.
Shown in the figure.

First, in step 31 in FIG. 4, for example, a drawing as shown in FIG. 6 (a) is read by the image reader 22 in FIG. 3 and stored as image data in the 24 file device.

In step 32, the image data obtained on the 24 file devices in step 31 by the central processing unit 25 by the algorithm shown in the literature 1 is used to (1) extract a black massive region and a boundary line (2) white (3) Extract point sequence data (4) Perform a series of steps to determine a symbol to recognize the symbol.

The recognition result is obtained as a list of types of recognition symbols and coordinate values of the symbols on the original drawing. FIG. 5 shows an example of this list.
NO in FIG. 5 is an element number of the array and serves as an identifier of each recognition result. X and Y of 42 and 43 in the figure are
This is the coordinate value of the symbol on the original drawing. In the type column of 44 in the figure, when a plurality of symbols are automatically recognized at the same time, it indicates which symbol is recognized, and the number corresponds to the symbol. For example, in this embodiment, 1 is a manhole symbol,
3 corresponds to a symbol of a number tag, and this number is hereinafter referred to as a symbol type number. FIG. 5 exists as an array on the temporary memory. For example, the sequence corresponding to FIG.
When the total number of recognized symbols is N, the array is an N-dimensional array, and each element of the array indicates the center coordinate of the symbol on the original drawing.
X, Y and kind for storing the kind number of the symbol
It is a structural variable with two integer type variables.

In step 33 in FIG. 4, the list of recognition results in the temporary memory obtained in step 32 is divided according to the symbol type number as shown in FIG. 7 and stored in the temporary memory. NO as the identifier of the symbol recognized in FIG. 7 is attached to each list.
That is, for a symbol identified as symbol type 1
NO is assigned in order from 1 and NO is also assigned in order from 1 for symbols identified as symbol type 2. Here, the seventh
Creating a separate list as shown in FIG. 5 separately from the list in FIG. 5 is a waste of memory, and FIG.
The list division shown in FIG. 7 is realized by adding the array shown in FIG. At this time, a variable for a flag for describing the correctness of the confirmation required in step 35 of FIG. 4 is also prepared, and a value 0 is set as an initial value (step 8).
Five). KNO in FIG. 8 is the same number as NO assigned to the divided list in FIG. FIG. 9 shows a flow for obtaining the arrangement shown in FIG. In order to assign a unique identification number for each type, for example, a counter is provided for sequentially assigning a symbol identifier from 1 for each type of symbol (step 81). For each recognition symbol, the value of the counter for that symbol type is set one after another (step 83).

A check sheet is created for each symbol type number using the list of recognition results for each symbol type number obtained above. First
The check sheet shown in the figure uses the list of recognition results corresponding to symbol type number 1 obtained up to step 33,
This is obtained by the algorithm shown in the figure.

Each step will be described in detail with reference to FIG. Step 91 and subsequent steps are executed for each symbol type. First, step 91
Then, the clipping range of the image that matches the symbol (width W, height H)
To determine. As shown in FIGS. 6B and 6C, the cutout size is variable depending on the type of the symbol. Step
In 92, the area (Lx × L) for creating a check sheet
y) is secured. Lx and Ly are Lx and Ly in FIG. 1, and are the width and length of the check sheet. Check sheet width Lx
Is set to the same value for all the symbols so as to be convenient for checking using the check sheet, even when the check sheet is output to paper and used for checking work. Ly is variable depending on the type of the symbol. When the total number of symbols recognized as the symbol is KIND_N and the number of cutout images that can be arranged on one line of the check sheet is L, it is obtained by the following equation. Here, L is a value calculated from the width W of one cutout image and the width Lx of the check sheet, and the symbol [] is a Gaussian symbol.

Lx = ([KIND_N ÷ L] +1) × H Steps 94 and 95 are executed for every recognition symbol.
In step 94, the image data on the file device 24 in FIG.
The image of the rectangular area of H is read. In step 95, the image data cut out in step 94 is arranged. As an example of the arrangement method, as shown in FIG. 1, the cutout images are arranged side by side in order, and when one line is satisfied, the cutout images are arranged in the next line. If the horizontal length of the check sheet is not an integral multiple of the cut-out image of the symbol, the part that is left halfway on the right side is supplemented with a blank image in step 97. In step 96, the last line is set to form rectangular image data as a whole.
If there is less than one line, the missing part is supplemented with a blank image. If there is no boundary between the cut-out images, it is difficult for the confirmer to see the boundary, so step 97 puts a grid of the size of the cut-out image in the entire rectangular image in which the cut-out images are arranged. Here, the check sheet is divided by a grid for each cut-out image that has been recognized. Therefore, if a grid is numbered as shown in FIG. 11, the grid number becomes an identifier of each cut-out image, and the grid number is Corresponds to NO. The image data obtained by the above procedure is a check sheet.

Next, using the check sheet created by the above method,
A method for confirming the recognized result will be described.

The check using the check sheet is performed by passing the image data of the check sheet in the temporary memory 23 in FIG. 3 to a person who outputs the check sheet, for example, a sheet of paper, to a checker, and using a writing tool such as a pen to recognize a symbol that is erroneously recognized. A method of marking a portion, a method of displaying a check sheet image on a CRT, and instructing a symbol which is erroneously recognized by the coordinate input device 29 in FIG. 3 can be considered.

Here, FIGS. 12 and 13 show the flow of the confirmation operation when the check sheet output on the paper is used. Twelfth
The figure shows the operation flow of the confirmer, and FIG. 13 shows the corresponding operation on the system side. The checker checks the symbols sequentially while looking at the created check sheet. If it is determined that the symbol is incorrectly recognized, a lattice number of the symbol is obtained. The lattice number is calculated by (L × X + Y) where, for example, when the cut-out image in which the error is confirmed is the X-th from the left and the Y-th from the top on the check sheet, the number of lines in the check sheet is L, and (L × X + Y). . The lattice number and the type number of the target symbol are input from an input device. On the other hand, the system is in the input waiting state shown in FIG. 13, and if there is an input from the confirmer, finds the corresponding symbol from the list of the recognition results held in the temporary memory and sets the value to 1 for the flag. (Step 124).

For example, a case will be described in which the recognition result of the manhole symbol is confirmed using the check sheet of FIG. When the check sheets are checked in order, it is found that the first symbol in the first check sheet of the manhole, the third symbol from the left, is incorrect (step 113). Since the grid number is calculated to be 3 (step 115), the symbol type number 1 and the grid number 3 are input to the system. The system that received the input
A search is made for a symbol having a type of 1 and a KNO of 3 from the array corresponding to the table in FIG. 8 in the temporary memory, and a fourth symbol is found as a corresponding one (steps 122 and 123), and the value 1 is set in the flag. (Step 124). When all the recognitions have been completed, the process proceeds to step 125, and the fourth symbol is not dumped to a file because the flag is 1, but is deleted from the table and corrected.

In the method of displaying a check sheet on a CRT and checking using the coordinate input device shown in FIG.
Steps 115 and 116 are replaced with the operation of pointing the coordinate input device to a point in the grid where a symbol determined to be incorrectly recognized on the check sheet is present. Further, by the operation of the system, a step for calculating the grid number from the coordinate values input from the coordinate input device is inserted between steps 122 and 123 in FIG. 13, as shown in FIG. Step 132 is newly added.

When the confirmation operation is completed, the system dumps the list of recognition results in the temporary memory to a file device, and ends the processing. When dumping to a file, check the flags of the elements in each array, and check the 5th
In the list in the figure, X and Y indicating the central coordinates of the symbol and the symbol type number recorded in the type column are stored. KN in Fig. 8
O and the flag are not stored because they are variables required only in the process of processing.

In addition to a processing device having a function of reading an image and automatically recognizing the image, the read image data and a list of the recognition results are passed through a medium such as an optical disk in the system shown in FIG.
Another embodiment of the present invention is to store a check sheet in a file device and to use the data to create a check sheet and confirm the result of automatic recognition.

〔The invention's effect〕

According to the present invention, when checking the recognition result, an image necessary for the check is cut out from the original drawing and a check sheet is created, so even a large drawing is limited as on a CRT. Work in a space is possible, and even when output on paper, for example, even if the original drawing is A1 size, the output of a check sheet in A4 size can save space on paper.

Regarding work efficiency, since the same symbols are arranged on the check sheet, finding symbols of images with different characteristics among the arranged symbols is an operation to find misrecognized symbols, This has the effect of simplifying the checking operation.

[Brief description of the drawings]

FIG. 1 is a diagram showing an example of the specification of a check sheet,
2A is a diagram showing a procedure for creating a check sheet, FIG. 2B is a diagram showing a procedure for confirming a recognition result using the check sheet, FIG. 3 is a configuration diagram of the system, and FIG. FIG. 5 is a diagram showing the overall procedure of the present invention for recognition, confirmation, and correction from drawing input, FIG. 5 is a diagram showing a list of recognition results obtained by automatic recognition, and FIG. FIGS. 6 (b) and 6 (c) are drawings showing images to be automatically recognized, and FIGS.
FIG. 7 is a conceptual diagram of a list in which a list of recognition results is divided according to symbol types, FIG. 8 is a diagram showing a format in which a divided list is realized, and FIG. 9 is a symbol for each recognition symbol. FIG. 10 is a diagram showing an algorithm for assigning a unique identifier for each type, FIG. 10 is a diagram showing an algorithm for creating a check sheet, FIG. 11 is a diagram showing a relationship between cutout images arranged on a check sheet and lattice numbers, FIG. 12 is a diagram showing a procedure of a confirmer who confirms a recognition result using a check sheet output on paper, FIG. 13 is a diagram showing a procedure of a system corresponding to the procedure of FIG. 12, FIG. 14 is a diagram showing the procedure of the system when confirming the recognition result on the check sheet displayed on the CRT. 22 image reader, 23 temporary memory, 24 file device, 25 processing device, 28 display device, 29 coordinate input device.

──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Sei Matsushima 1-280 Higashi Koikebo, Kokubunji-shi, Tokyo Inside Central Research Laboratory, Hitachi, Ltd. (56) References JP-A-59-206976 (JP, A) JP-A-63 -233465 (JP, A) JP-A-56-19168 (JP, A) JP-A-60-111490 (JP, A) (58) Fields investigated (Int. Cl. ⁶ , DB name) G06F 17/50 G06T 7/00

Claims (4)

(57) [Claims] An image data representing a drawing or a document having a figure, a character, or a symbol, type data of a symbol recognized based on the image data, coordinate values on an original drawing, and data indicating correctness of the recognized symbol. A recognition result list having the following is prepared in the storage means, and a check sheet in which image data of a predetermined range including the coordinate values of the recognized symbols are arranged is output on a display screen, and an arbitrary configuration of the displayed check sheet is performed. A method for confirming a recognition result, comprising: instructing externally the image data of (1), and correcting data indicating correctness of a recognition signal corresponding to the instructed image data. 2. The method according to claim 1, wherein the image data of the drawing or the document is obtained by optically scanning, and the coordinate value held in the recognition result list is the center of the recognized symbol. A method for recognizing a recognition result characterized by being coordinates. 3. The method according to claim 1 or 2, wherein when the image data of a predetermined range is displayed side by side as the check sheet, the recognition symbol data is divided and listed for each type. A method of confirming recognition data, wherein images are cut out in the order of symbols appearing in a list for each type and are arranged and displayed on a screen. 4. A method according to claim 1, wherein said predetermined range is determined for each type of target symbol.